Gradient-boosted signals on real market data. Triple-barrier labels, purged K-Fold CV, walk-forward backtests. Verified by Deflated Sharpe and Probability of Backtest Overfitting.
4 challengers killed by the paired multi-seed gate (storm shield, PEAD, insider cluster, sector momentum). Only plain 6-month momentum on SP500 cleared the bar — and only after the survivorship-corrected re-gate confirmed it (PIT Sharpe 0.71, down from 1.04 on the naive panel).
Orders route as marketable-limit at NBBO mid ± 50 bps with extended_hours so the bot never queues for the open auction. A daily-loss + drawdown auto-flatten kill switch SELLS the entire book on -3% intraday or -10% peak drawdown. Conviction sizing × inverse-vol × ¼-Kelly, vol-targeted.
GitHub Actions silently dropped 6 scheduled runs in two weeks. Fixed via 3 different non-round-hour cron schedules + a Vercel-cron triggered workflow_dispatch + idempotent client_order_id hashing so a same-day double fire dedupes at Alpaca.
Paired multi-seed t-test, stationary block bootstrap, n_trials-corrected Deflated Sharpe, Holm-Bonferroni multi-test. Every retraction is committed to retractions.json with the date, the p-value, and the git SHA. No silent kills.
Kill-switch · position caps · drawdown · max-positions.
Pending → submitted → filled / cancelled / rejected. Idempotent.
Top-K per universe, ordered by rank then confidence.
Reconnecting WS consumer, Redis pub/sub fan-out, 15 s heartbeats.
Equal-weight top-K, bps cost model, Sharpe + drawdown + DSR.
11 APIs keyed. Canonical Postgres. Zero synthetic paths.
No fake data.
No synthetic paths.
Real providers. Real markets.
Verified by deflated Sharpe.
Or it does not ship.
Measured from the commit log, not the pitch deck.
Real open positions on the Alpaca paper account, ranked by market value. Every 30 seconds from /api/signal-today. Zero synthetic paths.
